+12V and +5V regulation

Archived from groups: alt.comp.hardware.homebuilt (More info?)

I have a generic 400W ATX PSU whose +12V and +5V outputs vary
significantly under different load conditions. For example, at various
times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
good, this seemingly poor regulation is to be expected from the
design. Most PC PSUs regulate using a weighted average of the +5V and
+12V rails, so that when the +12V rail rises, the +5V rail falls to
maintain a constant average.

I'm considering modifying this PSU so that it regulates the +5V output
and allows the +12V rail to find its own level. This is easily done by
changing a few resistors.

For example, in the following PSU one could remove R25, and replace
R20 and R21 with a single 5K6 resistor:
http://www.pavouk.comp.cz/hw/en_atxps.html

This would guarantee a regulated +5V rail that was unaffected by
changes in the +12V load. Unfortunately the penalty would be that the
+12V rail may at times rise as high as 13V.

I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
+13V supply be a serious problem?


- Franc Zabkar
--
Please remove one 's' from my address when replying by email.
28 answers Last reply
More about regulation
  1. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Wed, 09 Mar 2005 20:15:28 +1100, Franc Zabkar
    <fzabkar@optussnet.com.au> wrote:

    >I have a generic 400W ATX PSU whose +12V and +5V outputs vary
    >significantly under different load conditions. For example, at various
    >times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
    >good, this seemingly poor regulation is to be expected from the
    >design. Most PC PSUs regulate using a weighted average of the +5V and
    >+12V rails, so that when the +12V rail rises, the +5V rail falls to
    >maintain a constant average.
    >
    >I'm considering modifying this PSU so that it regulates the +5V output
    >and allows the +12V rail to find its own level. This is easily done by
    >changing a few resistors.
    >
    >For example, in the following PSU one could remove R25, and replace
    >R20 and R21 with a single 5K6 resistor:
    > http://www.pavouk.comp.cz/hw/en_atxps.html
    >
    >This would guarantee a regulated +5V rail that was unaffected by
    >changes in the +12V load. Unfortunately the penalty would be that the
    >+12V rail may at times rise as high as 13V.

    .... or with modern systems using so much 12V power, a 12V
    rail that dips quite a bit. You don't mention the
    application of this modified PSU. If it's to be powering a
    *typical* PC then perhaps starting out with
    better-than-generic-400W would be desirable? Just wondering
    if the result would be worth the time spent, unless you only
    needed a switching 5V supply for other project(s).


    >
    >I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
    >+13V supply be a serious problem?

    In a recent thread someone had mentioned that one of the
    (suspected) failure modes of Maxtor HDDs were a couple of
    small 8 pin (SO-8) chips that were overheating. I meant to
    take a look at some Maxtor drives to determine what those
    might've been but suspect power regulators, maybe linear.
    If so, raising the 12V level might raise the thermals and
    tend to be more problematic for such drive designs... which
    could easily be more common than only (certain) Maxtor
    models.

    As for other 12V, it seems most CPU VRM circuits would work
    as well at a fairly high input variance, though raising 12V
    would tend to make other non-regulated devices (fans, some
    LED & misc case circuits) run a little
    hotter/brighter/louder/etc... which probably won't matter
    given the low % change.
  2. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    Franc Zabkar wrote:
    > I have a generic 400W ATX PSU whose +12V and +5V
    > outputs vary significantly under different load conditions.

    > Most PC PSUs regulate using a weighted average of the
    > +5V and +12V rails, so that when the +12V rail rises,
    > the +5V rail falls to maintain a constant average.
    >
    > I'm considering modifying this PSU so that it regulates
    > the +5V output and allows the +12V rail to find its own level.

    > This would guarantee a regulated +5V rail that was
    > unaffected by changes in the +12V load. Unfortunately
    > the penalty would be that the +12V rail may at times
    > rise as high as 13V.

    13V won't hurt anything because HDs don't use linear regulators except
    possibly for low power logic circuitry, and I'd expect them to run off
    the 5V anyway. The small chips someone referred to are most likely
    power drivers for the motor and are frequency and duty cycle modulated.

    Cheap power supplies often have badly designed output chokes that don't
    help the cross regulation as much as they should.
  3. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    Franc Zabkar wrote:
    > On Thu, 10 Mar 2005 11:28:27 +1300, "PC" <abuse@local.host> put
    finger
    > to keyboard and composed:

    > > I have found the biggest problem with PC supplies at the
    > > moment is the Chinese Capacitor problem.

    > My ATX PSU has had very little usage, but I'll check the caps anyway.

    My 2-3 year-old 300W Antec ran a Socket 7 AT system that drew only 60W
    from the +5V and +12V rails, but a couple of electrolytic capacitors on
    the +12V rail bulged anyway, despite powering only one HD and one
    CD-RW. At first, only one of these caps tested bad with the Bob Parker
    ESR meter (out of circuit because the caps were in parallel), but a
    week later a second one bulged and showed high ESR. Both were the
    Fuhjyyu brand, one purple, one black.
  4. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    Franc Zabkar wrote:

    > Just out of curiosity, would you please indulge me by testing your
    own
    > PSU? You should be able to use MBM5 for this purpose, assuming your
    > motherboard supports hardware monitoring. In particular I'd like to
    > know what happens if the +5V rail is heavily loaded while the +12V
    > rail is lightly loaded.

    I tested either a 1.6 GHz Duron, Athlon XP1800+, or 1.7 GHz Celeron in
    an ECS mobo that ran the CPU from the +5V rail, and the +5V current
    ranged from about 4.5-9A. A 7200 RPM HD, CPU fan, PSU fan, and CD-RW
    burner were the only loads on the +12V, and the +12V rail changed by no
    more than 0.1V whether the HD and CD-RW were at idle (probably only
    0.5A) or when a CD-RW disk was burned at either 4x or 8x speed (about
    1.6A). The PSU was a 300W Powmax but one of the older ones that was
    built decently, like a better Raidmax.
  5. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Wed, 09 Mar 2005 15:31:27 GMT, kony <spam@spam.com> put finger to
    keyboard and composed:

    >On Wed, 09 Mar 2005 20:15:28 +1100, Franc Zabkar
    ><fzabkar@optussnet.com.au> wrote:
    >
    >>I have a generic 400W ATX PSU whose +12V and +5V outputs vary
    >>significantly under different load conditions. For example, at various
    >>times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
    >>good, this seemingly poor regulation is to be expected from the
    >>design. Most PC PSUs regulate using a weighted average of the +5V and
    >>+12V rails, so that when the +12V rail rises, the +5V rail falls to
    >>maintain a constant average.
    >>
    >>I'm considering modifying this PSU so that it regulates the +5V output
    >>and allows the +12V rail to find its own level. This is easily done by
    >>changing a few resistors.
    >>
    >>For example, in the following PSU one could remove R25, and replace
    >>R20 and R21 with a single 5K6 resistor:
    >> http://www.pavouk.comp.cz/hw/en_atxps.html
    >>
    >>This would guarantee a regulated +5V rail that was unaffected by
    >>changes in the +12V load. Unfortunately the penalty would be that the
    >>+12V rail may at times rise as high as 13V.
    >
    >... or with modern systems using so much 12V power, a 12V
    >rail that dips quite a bit.

    .... which would result in a higher +5V rail in an unmodified PSU.

    >You don't mention the
    >application of this modified PSU.

    Athlon XP 2500+, 512MB DDR SDRAM, Xabre graphics card, 120MB HD, DVD
    ROM, DVD burner. Measured power consumption on AC side is ~160W.

    >If it's to be powering a
    >*typical* PC then perhaps starting out with
    >better-than-generic-400W would be desirable? Just wondering
    >if the result would be worth the time spent, unless you only
    >needed a switching 5V supply for other project(s).

    During my minicomputer days, the +5V rail was deemed of utmost
    importance as it powered all the logic. In contrast, the PC platform
    seems rather lax in this respect. In fact the only PC PSU I've
    encountered that has independent regulation for each of the major
    rails is Antec's Truepower. If a "better-than-generic-400W" PSU
    regulates the same way as a generic, then I can't see how it would be
    any more accurate. I'd be interested to know how other PSUs perform
    under varying load conditions. For example, what are the +5V and +12V
    readings when the HDs and DVD ROMs have spun down, as opposed to when
    they are all spinning and seeking? How are the voltage readings
    affected by CPU load, by multimedia apps, graphics card, etc?

    Do you have any numbers for your own PSU? You may be surprised.

    >>I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
    >>+13V supply be a serious problem?
    >
    >In a recent thread someone had mentioned that one of the
    >(suspected) failure modes of Maxtor HDDs were a couple of
    >small 8 pin (SO-8) chips that were overheating. I meant to
    >take a look at some Maxtor drives to determine what those
    >might've been but suspect power regulators, maybe linear.
    >If so, raising the 12V level might raise the thermals and
    >tend to be more problematic for such drive designs... which
    >could easily be more common than only (certain) Maxtor
    >models.

    Yes, that's my concern also, ie whether HDs and DVDs still use linear
    power techniques.

    >As for other 12V, it seems most CPU VRM circuits would work
    >as well at a fairly high input variance, though raising 12V
    >would tend to make other non-regulated devices (fans, some
    >LED & misc case circuits) run a little
    >hotter/brighter/louder/etc... which probably won't matter
    >given the low % change.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  6. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    "Franc Zabkar" <fzabkar@optussnet.com.au> wrote in message
    news:7lat2119n98886049g609em9btisre2t8l@4ax.com...
    >I have a generic 400W ATX PSU whose +12V and +5V outputs vary
    > significantly under different load conditions. For example, at various
    > times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
    > good, this seemingly poor regulation is to be expected from the
    > design. Most PC PSUs regulate using a weighted average of the +5V and
    > +12V rails, so that when the +12V rail rises, the +5V rail falls to
    > maintain a constant average.
    >
    > I'm considering modifying this PSU so that it regulates the +5V output
    > and allows the +12V rail to find its own level. This is easily done by
    > changing a few resistors.
    >
    > For example, in the following PSU one could remove R25, and replace
    > R20 and R21 with a single 5K6 resistor:
    > http://www.pavouk.comp.cz/hw/en_atxps.html
    >
    > This would guarantee a regulated +5V rail that was unaffected by
    > changes in the +12V load. Unfortunately the penalty would be that the
    > +12V rail may at times rise as high as 13V.
    >
    > I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
    > +13V supply be a serious problem?
    >
    >
    > - Franc Zabkar
    > --
    > Please remove one 's' from my address when replying by email.

    Franc

    3 thoughts come to mind

    If you realy do want to waste time on an apparent poorly regulated Power
    Supply then I strongly suggest you build a suitable (variable?) dummy load
    and connect your modified power supply to this during your 'experiments'.
    You don't want to risk blowing the rest of your PC if the power supply runs
    away and 'overvolts'

    For all the time and effort involved in redesigning someone else's work it
    will be better to just go buy a tighter regulated Power Supply.

    Is the circuit you give exactly the same make/model as yours? If not I would
    suggest yours will be vastly different.

    Cheers
    Paul.
  7. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    "Franc Zabkar" <fzabkar@optussnet.com.au> wrote in message
    news:3hqu21p28lt46hi60cqlgs355re0rpa024@4ax.com...
    > On Wed, 09 Mar 2005 15:31:27 GMT, kony <spam@spam.com> put finger to
    > keyboard and composed:
    >
    >>On Wed, 09 Mar 2005 20:15:28 +1100, Franc Zabkar
    >><fzabkar@optussnet.com.au> wrote:
    >>
    >>>I have a generic 400W ATX PSU whose +12V and +5V outputs vary
    >>>significantly under different load conditions. For example, at various
    >>>times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
    >>>good, this seemingly poor regulation is to be expected from the
    >>>design. Most PC PSUs regulate using a weighted average of the +5V and
    >>>+12V rails, so that when the +12V rail rises, the +5V rail falls to
    >>>maintain a constant average.
    >>>
    >>>I'm considering modifying this PSU so that it regulates the +5V output
    >>>and allows the +12V rail to find its own level. This is easily done by
    >>>changing a few resistors.
    >>>
    >>>For example, in the following PSU one could remove R25, and replace
    >>>R20 and R21 with a single 5K6 resistor:
    >>> http://www.pavouk.comp.cz/hw/en_atxps.html
    >>>
    >>>This would guarantee a regulated +5V rail that was unaffected by
    >>>changes in the +12V load. Unfortunately the penalty would be that the
    >>>+12V rail may at times rise as high as 13V.
    >>
    >>... or with modern systems using so much 12V power, a 12V
    >>rail that dips quite a bit.
    >
    > ... which would result in a higher +5V rail in an unmodified PSU.
    >
    >>You don't mention the
    >>application of this modified PSU.
    >
    > Athlon XP 2500+, 512MB DDR SDRAM, Xabre graphics card, 120MB HD, DVD
    > ROM, DVD burner. Measured power consumption on AC side is ~160W.
    >
    >>If it's to be powering a
    >>*typical* PC then perhaps starting out with
    >>better-than-generic-400W would be desirable? Just wondering
    >>if the result would be worth the time spent, unless you only
    >>needed a switching 5V supply for other project(s).
    >
    > During my minicomputer days, the +5V rail was deemed of utmost
    > importance as it powered all the logic. In contrast, the PC platform
    > seems rather lax in this respect. In fact the only PC PSU I've
    > encountered that has independent regulation for each of the major
    > rails is Antec's Truepower. If a "better-than-generic-400W" PSU
    > regulates the same way as a generic, then I can't see how it would be
    > any more accurate. I'd be interested to know how other PSUs perform
    > under varying load conditions. For example, what are the +5V and +12V
    > readings when the HDs and DVD ROMs have spun down, as opposed to when
    > they are all spinning and seeking? How are the voltage readings
    > affected by CPU load, by multimedia apps, graphics card, etc?
    >
    > Do you have any numbers for your own PSU? You may be surprised.
    >
    >>>I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
    >>>+13V supply be a serious problem?
    >>
    >>In a recent thread someone had mentioned that one of the
    >>(suspected) failure modes of Maxtor HDDs were a couple of
    >>small 8 pin (SO-8) chips that were overheating. I meant to
    >>take a look at some Maxtor drives to determine what those
    >>might've been but suspect power regulators, maybe linear.
    >>If so, raising the 12V level might raise the thermals and
    >>tend to be more problematic for such drive designs... which
    >>could easily be more common than only (certain) Maxtor
    >>models.
    >
    > Yes, that's my concern also, ie whether HDs and DVDs still use linear
    > power techniques.
    >
    >>As for other 12V, it seems most CPU VRM circuits would work
    >>as well at a fairly high input variance, though raising 12V
    >>would tend to make other non-regulated devices (fans, some
    >>LED & misc case circuits) run a little
    >>hotter/brighter/louder/etc... which probably won't matter
    >>given the low % change.
    >
    >
    > - Franc Zabkar
    > --
    > Please remove one 's' from my address when replying by email.

    Franc

    Have posted other thoughts to your original post. In it I suggested a
    variable dummy load for experimenting.
    This may be the better way to 'test' the supplies regulation.

    For my part while having built supplies (linear not switch mode) with
    regulation tighter than a fishes a... I have found the biggest problem with
    PC supplies at the moment is the Chinese Capacitor problem.
    i.e. on a meter they will show good regulation well within 5%, but when you
    look at the rails with a scope the noise can exceed 2 volts, This 'noise' is
    switching transients from inside the supply that are not removed by the Lo
    ESR caps in the supply. The caps of course invariably turn out to be some of
    the 'Chinese' one's suffering from the pirate electrolyte problem.

    Just another 2c worth

    Cheers
    Paul.
  8. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <7lat2119n98886049g609em9btisre2t8l@4ax.com>, Franc Zabkar
    says...
    > I have a generic 400W ATX PSU whose +12V and +5V outputs vary
    > significantly under different load conditions. For example, at various
    > times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
    > good, this seemingly poor regulation is to be expected from the
    > design.

    Are you using a multimeter or the info the motherboard reports?


    --
    Conor

    An imperfect plan executed violently is far superior to a perfect plan.
    -- George Patton
  9. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Thu, 10 Mar 2005 11:15:07 +1300, "PC" <abuse@local.host> put finger
    to keyboard and composed:

    >
    >"Franc Zabkar" <fzabkar@optussnet.com.au> wrote in message
    >news:7lat2119n98886049g609em9btisre2t8l@4ax.com...
    >>I have a generic 400W ATX PSU whose +12V and +5V outputs vary
    >> significantly under different load conditions. For example, at various
    >> times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
    >> good, this seemingly poor regulation is to be expected from the
    >> design. Most PC PSUs regulate using a weighted average of the +5V and
    >> +12V rails, so that when the +12V rail rises, the +5V rail falls to
    >> maintain a constant average.
    >>
    >> I'm considering modifying this PSU so that it regulates the +5V output
    >> and allows the +12V rail to find its own level. This is easily done by
    >> changing a few resistors.
    >>
    >> For example, in the following PSU one could remove R25, and replace
    >> R20 and R21 with a single 5K6 resistor:
    >> http://www.pavouk.comp.cz/hw/en_atxps.html
    >>
    >> This would guarantee a regulated +5V rail that was unaffected by
    >> changes in the +12V load. Unfortunately the penalty would be that the
    >> +12V rail may at times rise as high as 13V.
    >>
    >> I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
    >> +13V supply be a serious problem?
    >>
    >>
    >> - Franc Zabkar
    >> --
    >> Please remove one 's' from my address when replying by email.
    >
    >Franc
    >
    >3 thoughts come to mind
    >
    >If you realy do want to waste time on an apparent poorly regulated Power
    >Supply then I strongly suggest you build a suitable (variable?) dummy load
    >and connect your modified power supply to this during your 'experiments'.
    >You don't want to risk blowing the rest of your PC if the power supply runs
    >away and 'overvolts'
    >
    >For all the time and effort involved in redesigning someone else's work it
    >will be better to just go buy a tighter regulated Power Supply.

    Apart from Antec's TruePower PSU, how can I be sure that any branded
    PSU will regulate any better than my generic?

    Just out of curiosity, would you please indulge me by testing your own
    PSU? You should be able to use MBM5 for this purpose, assuming your
    motherboard supports hardware monitoring. In particular I'd like to
    know what happens if the +5V rail is heavily loaded while the +12V
    rail is lightly loaded.

    >Is the circuit you give exactly the same make/model as yours? If not I would
    >suggest yours will be vastly different.

    I'd be surprised if that were the case. I haven't worked on many ATX
    PSUs, but I have repaired and modified several AT PSUs (eg 12V ->
    13.8V, 5V -> 6V). They all seemed to follow similar design principles,
    at least in the area of interest. As for PSUs in general, I've built a
    few linears, and I've repaired probably one thousand PSUs of all
    kinds, ranging from 1000W switchers to sub-1W linears, in all kinds of
    equipment, from minicomputers to toasters.

    >Cheers
    >Paul.
    >


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  10. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On 9 Mar 2005 16:06:28 -0800, "Dick Cheney" <dick_cheney@my-deja.com>
    put finger to keyboard and composed:

    >
    >Franc Zabkar wrote:
    >> I have a generic 400W ATX PSU whose +12V and +5V
    >> outputs vary significantly under different load conditions.
    >
    >> Most PC PSUs regulate using a weighted average of the
    >> +5V and +12V rails, so that when the +12V rail rises,
    >> the +5V rail falls to maintain a constant average.
    >>
    >> I'm considering modifying this PSU so that it regulates
    >> the +5V output and allows the +12V rail to find its own level.
    >
    >> This would guarantee a regulated +5V rail that was
    >> unaffected by changes in the +12V load. Unfortunately
    >> the penalty would be that the +12V rail may at times
    >> rise as high as 13V.
    >
    >13V won't hurt anything because HDs don't use linear regulators except
    >possibly for low power logic circuitry, and I'd expect them to run off
    >the 5V anyway. The small chips someone referred to are most likely
    >power drivers for the motor and are frequency and duty cycle modulated.

    That's what I wanted to hear. I would think the voice coil amp would
    still be linear, though.

    >Cheap power supplies often have badly designed output chokes that don't
    >help the cross regulation as much as they should.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  11. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Thu, 10 Mar 2005 11:28:27 +1300, "PC" <abuse@local.host> put finger
    to keyboard and composed:

    >For my part while having built supplies (linear not switch mode) with
    >regulation tighter than a fishes a... I have found the biggest problem with
    >PC supplies at the moment is the Chinese Capacitor problem.

    I suspect these caps may have found their way into some high-end
    equipment as well.

    >i.e. on a meter they will show good regulation well within 5%, but when you
    >look at the rails with a scope the noise can exceed 2 volts, This 'noise' is
    >switching transients from inside the supply that are not removed by the Lo
    >ESR caps in the supply. The caps of course invariably turn out to be some of
    >the 'Chinese' one's suffering from the pirate electrolyte problem.

    My ATX PSU has had very little usage, but I'll check the caps anyway.
    I have an ESR meter (Bob Parker's design, sold in kit form) - it's the
    most useful item in my toolkit. It comes highly recommended by the
    folks at sci.electronics.repair. It can test caps in circuit.

    >Just another 2c worth
    >
    >Cheers
    >Paul.

    The "regulation" formula for the PSU at
    http://www.pavouk.comp.cz/hw/en_atxps.html is:

    (0.327 * V5) + (0.0678 * V12) = Vref / 2 = 2.50V (nominal)

    The corresponding relationship for my own PSU (based on external
    measurements) appears to be:

    (0.2654 * V5) + (0.09436 * V12) = 2.50V

    IMO, this intentional interdependence between the two rails is
    bizarre. AFAICS, the same design philosophy is used in branded PSUs.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  12. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <2qqv21lo7sosa0j4bo24cd2hiis86o8gcu@4ax.com>, Franc Zabkar
    says...

    > Just out of curiosity, would you please indulge me by testing your own
    > PSU? You should be able to use MBM5 for this purpose, assuming your
    > motherboard supports hardware monitoring. In particular I'd like to
    > know what happens if the +5V rail is heavily loaded while the +12V
    > rail is lightly loaded.
    >
    ROFLMFAO...

    Let me get this right....

    You're using the readings the motherboard reports? Jesus...


    --
    Conor

    An imperfect plan executed violently is far superior to a perfect plan.
    -- George Patton
  13. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Thu, 10 Mar 2005 08:57:21 +1100, Franc Zabkar
    <fzabkar@optussnet.com.au> wrote:


    >>If it's to be powering a
    >>*typical* PC then perhaps starting out with
    >>better-than-generic-400W would be desirable? Just wondering
    >>if the result would be worth the time spent, unless you only
    >>needed a switching 5V supply for other project(s).
    >
    >During my minicomputer days, the +5V rail was deemed of utmost
    >importance as it powered all the logic. In contrast, the PC platform
    >seems rather lax in this respect.

    The primary chipsets have that 5V stepped down now, though I
    don't know what else still relys on 5V as it could certainly
    vary per board.

    >In fact the only PC PSU I've
    >encountered that has independent regulation for each of the major
    >rails is Antec's Truepower.

    Is it always a positive thing though? It would seem to
    inherantly limit the current on any rail. AFAIK most of the
    major components now use one form or another of step down so
    to a certain extent it may not matter if the voltage is
    exactly right as much as the noise.


    >If a "better-than-generic-400W" PSU
    >regulates the same way as a generic, then I can't see how it would be
    >any more accurate. I'd be interested to know how other PSUs perform
    >under varying load conditions. For example, what are the +5V and +12V
    >readings when the HDs and DVD ROMs have spun down, as opposed to when
    >they are all spinning and seeking? How are the voltage readings
    >affected by CPU load, by multimedia apps, graphics card, etc?

    If no weight was placed on 12V, how would there be any way
    to determine under/overvoltage in such a budget-optimized
    design? Even if it's not technically as necessary to keep
    12V in tight tolernaces, it's still in the ATX spec, IIRC.


    >
    >Do you have any numbers for your own PSU? You may be surprised.

    Though I don't always trust motherboard monitoring software,
    I do often have it running on various systems, pariticularly
    those performing aux. functions that wouldn't be (otherwise)
    monitored as closely. Some of them do spin-down drives and
    some are mostly idle but others often under extended full
    load... voltages don't get too far out of line. Lowest of
    any system here (AFAIK) is a reading of 11.5 on 12V, though
    that's ~ 0.25V lower than it reads at the connector with a
    DMM.


    >
    >>>I realise that HDs (and PSUs) are rated for 12V +/-5%, but would a
    >>>+13V supply be a serious problem?
    >>
    >>In a recent thread someone had mentioned that one of the
    >>(suspected) failure modes of Maxtor HDDs were a couple of
    >>small 8 pin (SO-8) chips that were overheating. I meant to
    >>take a look at some Maxtor drives to determine what those
    >>might've been but suspect power regulators, maybe linear.
    >>If so, raising the 12V level might raise the thermals and
    >>tend to be more problematic for such drive designs... which
    >>could easily be more common than only (certain) Maxtor
    >>models.
    >
    >Yes, that's my concern also, ie whether HDs and DVDs still use linear
    >power techniques.

    Having just had someone's CDRW drive open today, I counted:

    1 - 78L05
    2 - LM317
    1 - 7809

    (maybe another, I wasn't really trying to enumerate them at
    the time) So certainly there's the potential for linear
    regulators on at least CDRW drives. This particular drive
    was a MSI/Acer/et al. 40X. However I don't claim to have
    tested the linear regulators on *dead* optical drives but
    suspect they aren't one of the primary failure points, and
    none of these low amperage SO-8 like what was (potentially)
    mentioned on some HDDs except the 78L05, which was only a
    supply for the front headphone opamp, IIRC.
  14. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Thu, 10 Mar 2005 15:42:15 -0000, Conor <conor@conorturton.com> put
    finger to keyboard and composed:

    >In article <7lat2119n98886049g609em9btisre2t8l@4ax.com>, Franc Zabkar
    >says...
    >> I have a generic 400W ATX PSU whose +12V and +5V outputs vary
    >> significantly under different load conditions. For example, at various
    >> times I can measure 4.89/12.74 and 5.05/12.29. While this doesn't look
    >> good, this seemingly poor regulation is to be expected from the
    >> design.
    >
    >Are you using a multimeter or the info the motherboard reports?

    Initially I used my DMM, but I'm now relying on MBM (and BIOS). The
    motherboard reports readings which are consistently lower than the
    metered readings at the ATX connector by ~50mV. The resolution of the
    hardware monitor is 16mV per count, which gives resolutions of 27mV
    and 64mV for the +5V and +12V rails, respectively. Therefore a change
    from 12.74 to 12.29 (450mV) is significant. BTW, I've calibrated my
    DMM against a 5.000V +/- 0.001 reference based on Maxim's MAX6350.

    See the second half of this very simple circuit consisting of just one
    IC plus caps:
    http://www.siliconchip.com.au/cms/A_30853/article.html


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  15. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Thu, 10 Mar 2005 15:41:44 -0000, Conor <conor@conorturton.com> put
    finger to keyboard and composed:

    >In article <2qqv21lo7sosa0j4bo24cd2hiis86o8gcu@4ax.com>, Franc Zabkar
    >says...
    >
    >> Just out of curiosity, would you please indulge me by testing your own
    >> PSU? You should be able to use MBM5 for this purpose, assuming your
    >> motherboard supports hardware monitoring. In particular I'd like to
    >> know what happens if the +5V rail is heavily loaded while the +12V
    >> rail is lightly loaded.
    >>
    >ROFLMFAO...
    >
    >Let me get this right....
    >
    >You're using the readings the motherboard reports? Jesus...

    Once again you've ejaculated prematurely. See my other reply ...


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  16. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On 9 Mar 2005 23:44:04 -0800, "larry moe 'n curly"
    <larrymoencurly@my-deja.com> put finger to keyboard and composed:

    >
    >Franc Zabkar wrote:
    >> On Thu, 10 Mar 2005 11:28:27 +1300, "PC" <abuse@local.host> put
    >finger
    >> to keyboard and composed:
    >
    >> > I have found the biggest problem with PC supplies at the
    >> > moment is the Chinese Capacitor problem.
    >
    >> My ATX PSU has had very little usage, but I'll check the caps anyway.
    >
    >My 2-3 year-old 300W Antec ran a Socket 7 AT system that drew only 60W
    >from the +5V and +12V rails, but a couple of electrolytic capacitors on
    >the +12V rail bulged anyway, despite powering only one HD and one
    >CD-RW. At first, only one of these caps tested bad with the Bob Parker
    >ESR meter (out of circuit because the caps were in parallel), but a
    >week later a second one bulged and showed high ESR.

    So what does your experience with dud caps in a branded product tell
    you? ;-)

    >Both were the Fuhjyyu brand, one purple, one black.

    I thought you made that up until I found their website. :-)


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  17. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Fri, 11 Mar 2005 08:21:30 +1100, Franc Zabkar
    <fzabkar@optussnet.com.au> wrote:


    >>My 2-3 year-old 300W Antec ran a Socket 7 AT system that drew only 60W
    >>from the +5V and +12V rails, but a couple of electrolytic capacitors on
    >>the +12V rail bulged anyway, despite powering only one HD and one
    >>CD-RW. At first, only one of these caps tested bad with the Bob Parker
    >>ESR meter (out of circuit because the caps were in parallel), but a
    >>week later a second one bulged and showed high ESR.
    >
    >So what does your experience with dud caps in a branded product tell
    >you? ;-)
    >
    >>Both were the Fuhjyyu brand, one purple, one black.
    >
    >I thought you made that up until I found their website. :-)


    LOL.

    I wondered about the spelling but I have several of those
    sitting in a scap bin, though I don't recall ever seeing the
    purple ones. Offhand I vaguely recall they were used in
    some other major PSU brands, maybe Sparkle for awhile.
    Don't remember too many of them failing though and find it
    odd/rare that there were two paralleled in the circuit.
    More often it seems they'll stuff one before an inductor and
    one after.
  18. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <qba1311tlfh1crl3hc0ca1h150r4mkhtsb@4ax.com>, Franc Zabkar
    says...

    > Once again you've ejaculated prematurely. See my other reply ...
    >
    Actually I didn't. Get a newsreader that threads conversations
    properly.


    --
    Conor

    An imperfect plan executed violently is far superior to a perfect plan.
    -- George Patton
  19. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Fri, 11 Mar 2005 09:39:16 -0000, Conor
    <conor@conorturton.com> wrote:


    >I particularly like the following bit which describes your problem you
    >"think" you have:
    >
    >"Another interesting trend I noticed was that percent-variance on 13
    >out of 14 individual rails tested increased when the board underwent
    >load."
    >
    >"In fact, I wouldn't hesitate to bet that some people have been scared
    >into buying a new power supply just because of some low voltage
    >readings that were the fault of the motherboard rather than the
    >original power supply."


    That is true when someone relies ONLY on the motherboard
    software for a reading, but he appears to have already
    compared this to a meter reading so for this particular
    isolated system it is known what the variance is.
  20. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    kony wrote:
    > On Fri, 11 Mar 2005 09:39:16 -0000, Conor
    > <conor@conorturton.com> wrote:
    >
    >
    >
    >>I particularly like the following bit which describes your problem you
    >>"think" you have:
    >>
    >>"Another interesting trend I noticed was that percent-variance on 13
    >>out of 14 individual rails tested increased when the board underwent
    >>load."
    >>
    >>"In fact, I wouldn't hesitate to bet that some people have been scared
    >>into buying a new power supply just because of some low voltage
    >>readings that were the fault of the motherboard rather than the
    >>original power supply."
    >
    >
    >
    > That is true when someone relies ONLY on the motherboard
    > software for a reading, but he appears to have already
    > compared this to a meter reading so for this particular
    > isolated system it is known what the variance is.

    True, but I'll be interested to see his reply because that article suggests
    the offset is not constant over load.
  21. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <o683311hn8vebf4m18tgeuae0424cke0ki@4ax.com>, kony says...

    > That is true when someone relies ONLY on the motherboard
    > software for a reading, but he appears to have already
    > compared this to a meter reading so for this particular
    > isolated system it is known what the variance is.
    >
    Oh dear oh dear oh dear. He measured it with very little load on. The
    motherboard then reported a drop in voltage under a high load. The
    article points out the fact that the margin of error of the
    motherboards INCREASES as the load goes up.

    --
    Conor

    An imperfect plan executed violently is far superior to a perfect plan.
    -- George Patton
  22. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Fri, 11 Mar 2005 21:25:10 -0000, Conor
    <conor@conorturton.com> wrote:

    >In article <o683311hn8vebf4m18tgeuae0424cke0ki@4ax.com>, kony says...
    >
    >> That is true when someone relies ONLY on the motherboard
    >> software for a reading, but he appears to have already
    >> compared this to a meter reading so for this particular
    >> isolated system it is known what the variance is.
    >>
    >Oh dear oh dear oh dear. He measured it with very little load on. The
    >motherboard then reported a drop in voltage under a high load. The
    >article points out the fact that the margin of error of the
    >motherboards INCREASES as the load goes up.


    One issue not mentioned is whether the CPU uses 12V or 5V
    derived power. If 5V, the 12V load won't be changing very
    much at all (until drives spin down/up), there would be no
    "high load" relative to the constant load.
  23. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <1ti731lieuttct0a3uerv062i2i53cfu8v@4ax.com>, Franc Zabkar
    says...
    > Anyway I tested my system and came to the conclusion that my PSU is
    > very poor, at least as far as +5V/+12V regulation is concerned. MBM's
    > reporting appears to be reasonably close to the mark after allowing
    > for voltage drops in the PSU harness and on the motherboard itself.

    Once again you fail to take into account that the accuracy of the
    motherboard monitoring DECREASES with load.

    --
    Conor

    An imperfect plan executed violently is far superior to a perfect plan.
    -- George Patton
  24. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Thu, 10 Mar 2005 22:46:46 GMT, kony <spam@spam.com> put finger to
    keyboard and composed:

    >On Fri, 11 Mar 2005 08:21:30 +1100, Franc Zabkar
    ><fzabkar@optussnet.com.au> wrote:
    >
    >
    >>>My 2-3 year-old 300W Antec ran a Socket 7 AT system that drew only 60W
    >>>from the +5V and +12V rails, but a couple of electrolytic capacitors on
    >>>the +12V rail bulged anyway, despite powering only one HD and one
    >>>CD-RW. At first, only one of these caps tested bad with the Bob Parker
    >>>ESR meter (out of circuit because the caps were in parallel), but a
    >>>week later a second one bulged and showed high ESR.
    >>
    >>So what does your experience with dud caps in a branded product tell
    >>you? ;-)
    >>
    >>>Both were the Fuhjyyu brand, one purple, one black.
    >>
    >>I thought you made that up until I found their website. :-)
    >
    >
    >LOL.
    >
    >I wondered about the spelling but I have several of those
    >sitting in a scap bin, though I don't recall ever seeing the
    >purple ones.

    I don't know whether their current product range
    (http://www.fuhjyyu.com.tw/pro.htm) is representative of their past
    efforts, but the only purple/violet caps I see are the 85 degC general
    purpose SK series:
    http://www.fuhjyyu.com.tw/pro_sk-1.htm

    >Offhand I vaguely recall they were used in
    >some other major PSU brands, maybe Sparkle for awhile.
    >Don't remember too many of them failing though and find it
    >odd/rare that there were two paralleled in the circuit.
    >More often it seems they'll stuff one before an inductor and
    >one after.

    Yeah, I wondered about that. A continuity test with a DMM would
    suggest they were in parallel, whereas an ESR test (with Bob Parker's
    meter) would see the inductor as a relatively high impedance.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  25. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Sat, 12 Mar 2005 02:12:43 GMT, kony <spam@spam.com> put finger to
    keyboard and composed:

    >On Fri, 11 Mar 2005 21:25:10 -0000, Conor
    ><conor@conorturton.com> wrote:
    >
    >>In article <o683311hn8vebf4m18tgeuae0424cke0ki@4ax.com>, kony says...
    >>
    >>> That is true when someone relies ONLY on the motherboard
    >>> software for a reading, but he appears to have already
    >>> compared this to a meter reading so for this particular
    >>> isolated system it is known what the variance is.
    >>>
    >>Oh dear oh dear oh dear. He measured it with very little load on. The
    >>motherboard then reported a drop in voltage under a high load. The
    >>article points out the fact that the margin of error of the
    >>motherboards INCREASES as the load goes up.
    >
    >
    >One issue not mentioned is whether the CPU uses 12V or 5V
    >derived power. If 5V, the 12V load won't be changing very
    >much at all (until drives spin down/up), there would be no
    >"high load" relative to the constant load.

    The Vcore regulator uses the +5V rail. AFAICT, the +12V rail powers
    the fans and drives, and not much else.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  26. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    In article <fse5311b9mf5uu3r2r8uq6quaut8i8u55g@4ax.com>, Franc Zabkar
    says...

    > The Vcore regulator uses the +5V rail. AFAICT, the +12V rail powers
    > the fans and drives, and not much else.
    >
    And as the voltages are regulated down to 3.3V and lower, 5V being a
    little out isn't going to matter.


    --
    Conor

    An imperfect plan executed violently is far superior to a perfect plan.
    -- George Patton
  27. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Sun, 13 Mar 2005 13:32:24 -0000, Conor <conor@conorturton.com> put
    finger to keyboard and composed:

    >In article <fse5311b9mf5uu3r2r8uq6quaut8i8u55g@4ax.com>, Franc Zabkar
    >says...
    >
    >> The Vcore regulator uses the +5V rail. AFAICT, the +12V rail powers
    >> the fans and drives, and not much else.
    >>
    >And as the voltages are regulated down to 3.3V and lower, 5V being a
    >little out isn't going to matter.

    It definitely won't matter to the CPU, but there are other devices
    which depend on this rail. Even so, you may be right, it may not
    matter to them either. I come from the old school where every supply
    voltage could be trimmed. In fact, flaky intermittent problems could
    sometimes be narrowed down by misadjusting a particular supply rail.
    Overclockers also recognise that minor supply adjustments can
    sometimes improve the stability of their setup.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
  28. Archived from groups: alt.comp.hardware.homebuilt (More info?)

    On Sun, 13 Mar 2005 13:31:34 -0000, Conor <conor@conorturton.com> put
    finger to keyboard and composed:

    >In article <1ti731lieuttct0a3uerv062i2i53cfu8v@4ax.com>, Franc Zabkar
    >says...
    >> Anyway I tested my system and came to the conclusion that my PSU is
    >> very poor, at least as far as +5V/+12V regulation is concerned. MBM's
    >> reporting appears to be reasonably close to the mark after allowing
    >> for voltage drops in the PSU harness and on the motherboard itself.
    >
    >Once again you fail to take into account that the accuracy of the
    >motherboard monitoring DECREASES with load.

    If you are referring to ohmic losses in the conductors and in the
    motherboard power planes, then clearly this is evident in the results.
    I've even measured the voltage drops in the cable. I've also pointed
    out the positions of the various circuit blocks so that the reader
    could assess the potential for inaccuracies. For example, the ATX
    power connector and monitoring chip are at diametrically opposite ends
    of the PCB, which is potentially a worst case scenario for voltage
    measurements. However, even allowing for these voltage drops, the
    hardware monitor appears to be close to the mark.

    If you believe there is some factor other than ohmic loss which is
    causing inaccuracies, then what could it be?

    Anyway, regardless of inaccuracies, the article you referred to proves
    the point that it set out to test, ie that voltage measurements
    reported by motherboard monitoring software do not always accurately
    reflect the actual output of the PSU, nor can they. I've said as much
    in several prior discussions in Usenet. It is for precisely this
    reason that I've taken the time to assess the accuracy of my own
    setup. If I were the author of that test, I would have taken the next
    step and investigated *why* some results (eg AOpen AK79D) were way off
    the mark. In my case, I downloaded the datasheet and a reference
    circuit for my monitor IC and used the information to properly
    configure MBM.

    http://www.ite.com.tw/product_info/file/pc/IT8705F_V.0.4.1.pdf

    AOpen's AK79D appears to use Winbond's W83627HF IC:
    http://www.winbond.com/PDF/sheet/w83627hf.pdf

    This IC directly monitors the +5V rail via on-chip scaling resistors,
    so I would have thought that its accuracy would be better than mine
    (the IT8705 requires external resistors). Anyway, if that were my
    system, I'd be trying to find a reason for the 0.5V drop in the 5V
    rail. Perhaps the motherboard has poorly designed power planes??? Is
    the ATX power connector showing signs of discolouration? How does MBM
    report the same parameters?

    Having said all the above, note that my +12V rail, as measured by MBM,
    INCREASES when the +5V rail is loaded, even though the load on the
    +12V rail does not change. Clearly this points to a regulation problem
    in the PSU. I have since dismantled the unit and can confirm that it
    regulates by sensing a weighted average of its +5V and +12V outputs.
    This explains why the two rails move in opposite directions.


    - Franc Zabkar
    --
    Please remove one 's' from my address when replying by email.
Ask a new question

Read More

Homebuilt Power Supplies Systems